Sound Analysis

sound analysis screenshot

Seeing Sound, the Resonance Profile: The playing of a fine musician is the true measure of a violin, as can be heard in Music Player, but for deeper analysis we need consistent tests, and a way to record and display the results. The frequency response graphs or Spectra displayed here are derived from recordings of sound "radiation" tests of the Plowden, Willemotte, and Titian violins, using an "impact hammer" impulse system.

These tests do reveal key components of a violin's sound envelope. Other aspects have been more resistant to full testing, as much of what we hear involves complex psychoacoustic processes, how bowed notes start and stop (transient response), or elusive playability issues. These "intangibles" are the focus of ongoing research.

About the Graphs: To create these displays, the recorded sound samples were analyzed with a computer program, and the processed results displayed as a Spectrum, a graph which visually indicates the amount of sound radiated at each frequency (musical pitch). The various component frequencies are displayed progressively higher from left to right, and amplitude or loudness is displayed progressively louder from bottom to top. The frequencies are measured in hertz (cycles per second), and the loudness in decibels *(see note below)..

Impact Hammer Tests: For repeatability of these sound samples the violin was suspended in a holding apparatus, and the sound was generated using a tiny "impact hammer", which lightly tapped the bass corner of the bridge in a direction parallel with the plane of the top plate. This activates all the resonances of the instrument, for a consistent sonic "snapshot".

The measurement file for each violin consists of twelve individual bridge-tap readings, each recorded at a particular microphone position surrounding the violin The first position is directly in front of the bridge. For each successive one, the violin is rotated to the left by 30 degrees, (i.e.; over the post, then treble f-hole, treble rib, and so forth, with tap number six directly facing the back,) so that the twelve readings are equally spaced around the instrument. (See About the Sound-Test WAV Files below).

Using the "Bridge-Tap" Sound Player: Check the box next to the violin name to play the recorded series of bridge taps made with the impact hammer. While obviously unlike a played violin, the taps from each instrument are quite revealing and distinctive to the ear. Listening to the series of bridge taps, one can easily hear the sound variation from different positions on the violin, with tap #6 at the back noticeably quieter than tap #1 facing the front. The entire series of 12 taps were averaged and analyzed for this display. These sample tests are offered for demonstration only; Dr. Bissinger's highly calibrated Acoustic Scans (link????) of the project violins were done separately in a special anechoic chamber with a full array of microphones.

Using the Spectra Viewer: The viewer allows comparisons between instruments, and between display views as well. At least one violin and one view must be checked to see a spectra image. E.g., check two violins and check the detail view; or one violin with detail view and 1/3 octave. The graphs are displayed in colors-Plowden shades of blue; Willemotte shades of red, Titian shades of Green/yellow

The Display Views: The Spectra here were processed in a choice of three views:

  • Detail, which shows individual peaks, each indicating a strong resonance
  • 1/9 Octave, averaged in 1/9 of an octave sections, and smoothed for ease of visual comparison.
  • 1/3 Octave, averaged in larger 1/3 of an octave sections   even less detail but more revealing of overall trends.

About the Sound-Test WAV Files: Included in GET FILES are three stereo WAV files derived from sound radiation measurements of the Plowden, Titian, and Willemotte violins. Because this set of measurements was made under far-from-ideal conditions, using an early, uncalibrated version of Joseph Curtin's Impact Hammer Rig, they should not be taken as fully accurate measurements of the instruments. Rather, they can be considered samples of the sort of data generated by impact hammer measurements. (see Bissinger, Acoustic Scans, EXCEL 7.8mb)

  • These test recordings are stereo .wav files. The left channel of each .wav file is the microphone signal, and the right channel is the hammer signal.
  • The hammer tapped the bass corner of the bridge in a direction parallel with the plane of the top plate.
  • The cardioid microphone was positioned 37 cm from the central axis of the instrument, here defined as a line rising vertically through the endpin.
  • There are twelve individual readings in each file. Each is the complex average of four individual measurements taken at the same microphone position. This kind of averaging reduces random noise.

To process this data using a two-channel sound analysis program such as SpectraPlus, set the processor to do a "transfer function" between left and right channels (left/right). In essence the recorded hammer-tap signal (right channel) is subtracted from the recorded sound radiation (left channel) .This calculation equalizes any variation in the hammer taps, as the harder 'tap' yields an equivalently louder sound radiation. A transfer function analyzed for an individual microphone position yields a "frequency response function" representing the sound pressure at the microphone position, per unit force at the bridge. The average magnitude of all twelve positions gives an estimate of the total sound radiation.

*Note; In the spectrum graphs here, amplitude is plotted on the vertical scale using a decibel (dB) scale, which is logarithmic and roughly mimics our sense of loudness. Frequency, (or vibration cycles per second,) is what we would hear as musical pitch, and is measured in hertz (Hz) and plotted on the horizontal axis, again using a logarithmic scale, so that each octave is given equal space.